|
Øyvind Sture: Ambient occlusion shading of seismic data
Abstract
This programming project consists of two parts. The first part deals with creating a
file loader for volumetric seismic data that has been stored in the
standardized SEG Y format. SEG
Y data files are used to store geophysical data acquired from reflection
seismology. The second part deals with visualizing the loaded SEG Y volume using the shading model called ’local ambient
occlusion’. The renderings using the local ambient occlusion shading model is
compared with renderings using the de facto Blinn Phong shading model. The comparisons show that the novel
combination of seismic volumetric data with local ambient occlusion give
superior results when it comes to revealing critical details in the seismic
data.
Both the file loader and the shader are written as separate plugins for the
Volumeshop software which is an interactive system for direct volume
illustration. The programming is done on a Windows computer, using Visual C++
and OpenGL. The hardware consists of an NVIDIA Geforce 9600GT, 512Mb graphics
card. The shading is done with a naive and nonoptimized implementation of local
ambient occlusion.
Screenshots
Results
Figure 5
and 6 shows a test dataset specifically created with a wedge, corners, flat
areas and a hold to check that the ambient shading was calculated correctly. Figure
5 is rendered with Phong, and figure 6 is rendered with Local Ambient
Occlusion. The cut-out parts of the object is where the shading effect is most
visible.
Figure 7 shows the head dataset rendered with Phong shading. Figure 8 shows the
difference between 14 rays for each voxel, and 26 rays for each voxel. The
image rendered with 26 rays looks smoother than the image rendered with 14
rays. There are still some artifacts in the image, but one can see that more
rays improve the image quality. In the paper by Hernell et al. they use up to
128 rays to create images.
Figure 9 shows the difference when the seismic dataset is shaded with Phong and
with Local Ambient Occlusion, with 14 and 26 rays. The layers (so called
horizons) in the seismic dataset are more clearly visible when the Local
Ambient Occlusion model is used, and it is also easier to see where the dataset
are more sparse. When comparing the image rendered
with 14 rays and the image rendered with 26 rays, one can see that the top
layer is smoother in the image with 26 rays. Figure 10 shows a split image with
Phong shading on the left side and Local Ambient Occlusion on the right. Figure
11 shows a striped image with Phong shading and Local Ambient Occlusion. Figure
12 shows magnified selected areas of sections of figure 9. It can be seen that
local ambient occlusion yields less noisy images with more clearly delineated
layer structures.
Conclusion and problems
The
implementation on the GPU without the use of precalculation led to some problems. The rendering of the
objects are very slow, up to multiple seconds. One possible speedup would be to
store the computed light values when they were computed, or to do this in a precalculation part, since the Local Ambient Occlusion
model is independent of the view vector.
Another problem was that for large datasets, or when I tried with more rays,
overloading the graphics card with intense calculations led to frequent bluescreens, or artifacts in the
rendered images.
The use of a Local Ambient Occlusion model appears to have a big impact on
visualization of noisy datasets like seismic dataset.
Related work
Ambient occlusion of medical data is discussed in [4] and quick ambient occlusion rendering is covered in [7].
Ambient occlusion applied on molecular data can be seen in [5]. In [6] it is discussed a variation of the phong
shading model on seismic data by using a surface normal which is better suited for seismic data.
Resources
Bibliography
[1] http://www.volumeshop.org/
[2] http://www.seg.org/
[3] http://www.seg.org/SEGportalWEBproject/prod/SEG-Publications/Pub-
Technical-Standards/Documents/seg y rev1.pdf
[4] F. Hernell, P. Ljung and A. Ynnerman, Efficient Ambient and Emissive Tissue
Illumination using Local Occlusion in Multiresolution Volume Rendering,
In Eurographics/IEEE-VGTC Symposium on Volume Graphics (2007)
[5] http://qutemol.sourceforge.net/sidetoside/
[6] 3D Seismic Volume Rendering, Pedro Mário Silva, Marcos Machado, Marcelo Gattass, 8th International Congress of The Brazilian Geophysical Society, Rio de Janeiro, Brazil, 14-18 September 2003.
[7] Interactive Volume Rendering with Dynamic Ambient Occlusion and Color Bleeding. Timo Ropinski, Jennis Meyer-Spradow, Stefan Diepenbrock, Jörg Mensmann and Klaus Hinrichs. Eurographics 2008
[8] http://en.wikipedia.org/wiki/Blinn-Phong shading model
[9] http://www.cs.umu.se/~isak/Snippets/a2e.c
[10] http://en.wikipedia.org/wiki/Ascii
[11] http://en.wikipedia.org/wiki/EBCDIC
[12] http://en.wikipedia.org/wiki/Reflection seismology
[13] http://en.wikipedia.org/wiki/Endianness
Supervisor: Daniel Patel
Test data: here
Volumeshop Plugins: here
|